Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment Mapping
To efficiently use the finite wireless communication resource (radio spectrum), a Radio Environment Map (REM) is needed to monitor, analyse and provide rich awareness of spectrum activities in a radio propagating environment. REM shows radio coverage metrics in a geographical region. A REM construct...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10813343/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846099177599139840 |
|---|---|
| author | Oluwatobi Emmanuel Dare Kennedy Okokpujie Emmanuel Adetiba Olabode Idowu-Bismark Abdultaofeek Abayomi Raymond Jules Kala Emmanuel Owolabi Udeme Christopher Ukpong |
| author_facet | Oluwatobi Emmanuel Dare Kennedy Okokpujie Emmanuel Adetiba Olabode Idowu-Bismark Abdultaofeek Abayomi Raymond Jules Kala Emmanuel Owolabi Udeme Christopher Ukpong |
| author_sort | Oluwatobi Emmanuel Dare |
| collection | DOAJ |
| description | To efficiently use the finite wireless communication resource (radio spectrum), a Radio Environment Map (REM) is needed to monitor, analyse and provide rich awareness of spectrum activities in a radio propagating environment. REM shows radio coverage metrics in a geographical region. A REM construction model with few constraints and optimal performance is needed to better support cognitive radio for dynamic spectrum sharing (DSS) and other benefits of REM. This study aims to estimate fine-resolution REM from sparse radio signal strength measurement. In this study, we utilised conditional generative adversarial network (CGAN) to create a television spectrum radio environment map in order to improve cognitive television white space (TVWS) radio performance in real-time propagation environments. Measurement campaign was carried out to acquire a TV-band (470-862MHz) radio frequency and geographical dataset at Covenant University, Ota, Nigeria. A preprocessing procedure which was implemented with Python script was employed to group the dataset using Nigerian Communications Commission TV spectrum channel spacing and to create incomplete spectrograms for 49 channels. Xgboost, SVM, and kriging variogram models were explored to generate ground truth datasets for the CGAN model training, and the best algorithm was employed. A CGAN REM model was developed using U-Net as a generator and PatchGan as a discriminator. The U-Net generator is a 3-channel input, 16-layer architecture while the PatchGan discriminator is a 6-channel input, 7-layer architecture. The model performance was evaluated using mean square error (MSE) and mean absolute error (MAE). 12 different experiments were carried out varying the training parameters of the CGAN architecture to obtain an optimal model. The achieved root mean square error (RMSE) is 0.1145dBm and MAE is 0.0820dBm, which shows the deviation between the ground truth and the generated REM. This low deviation means that the proposed CGAN REM model possesses an improved accuracy in predicting the spectrum activities within the television spectrum which is considered appropriate for DSS technology. This study also revealed that 41 channels within TV-band in Covenant University are totally unoccupied. |
| format | Article |
| id | doaj-art-8226b55d2bd54bb6a8cfda51c30a2798 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-8226b55d2bd54bb6a8cfda51c30a27982025-01-01T00:01:27ZengIEEEIEEE Access2169-35362024-01-011219763219764410.1109/ACCESS.2024.352199810813343Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment MappingOluwatobi Emmanuel Dare0https://orcid.org/0009-0002-4122-947XKennedy Okokpujie1https://orcid.org/0000-0002-7594-276XEmmanuel Adetiba2https://orcid.org/0000-0001-9227-7389Olabode Idowu-Bismark3Abdultaofeek Abayomi4https://orcid.org/0000-0003-3129-5246Raymond Jules Kala5https://orcid.org/0000-0001-6265-389XEmmanuel Owolabi6Udeme Christopher Ukpong7https://orcid.org/0009-0005-4901-8601Department of Electrical and Information Engineering, College of Engineering, Covenant University, Ota, NigeriaDepartment of Electrical and Information Engineering, College of Engineering, Covenant University, Ota, NigeriaDepartment of Electrical and Information Engineering, College of Engineering, Covenant University, Ota, NigeriaDepartment of Electrical and Information Engineering, College of Engineering, Covenant University, Ota, NigeriaIASRG, Summit University, Offa, Kwara, NigeriaSTEM Faculty, International University of Grand-Bassam, Grand-Bassam, Côte d’IvoireDepartment of Science, Mathematics and Technology Education, University of Pretoria, Pretoria, South AfricaDepartment of Electrical and Information Engineering, College of Engineering, Covenant University, Ota, NigeriaTo efficiently use the finite wireless communication resource (radio spectrum), a Radio Environment Map (REM) is needed to monitor, analyse and provide rich awareness of spectrum activities in a radio propagating environment. REM shows radio coverage metrics in a geographical region. A REM construction model with few constraints and optimal performance is needed to better support cognitive radio for dynamic spectrum sharing (DSS) and other benefits of REM. This study aims to estimate fine-resolution REM from sparse radio signal strength measurement. In this study, we utilised conditional generative adversarial network (CGAN) to create a television spectrum radio environment map in order to improve cognitive television white space (TVWS) radio performance in real-time propagation environments. Measurement campaign was carried out to acquire a TV-band (470-862MHz) radio frequency and geographical dataset at Covenant University, Ota, Nigeria. A preprocessing procedure which was implemented with Python script was employed to group the dataset using Nigerian Communications Commission TV spectrum channel spacing and to create incomplete spectrograms for 49 channels. Xgboost, SVM, and kriging variogram models were explored to generate ground truth datasets for the CGAN model training, and the best algorithm was employed. A CGAN REM model was developed using U-Net as a generator and PatchGan as a discriminator. The U-Net generator is a 3-channel input, 16-layer architecture while the PatchGan discriminator is a 6-channel input, 7-layer architecture. The model performance was evaluated using mean square error (MSE) and mean absolute error (MAE). 12 different experiments were carried out varying the training parameters of the CGAN architecture to obtain an optimal model. The achieved root mean square error (RMSE) is 0.1145dBm and MAE is 0.0820dBm, which shows the deviation between the ground truth and the generated REM. This low deviation means that the proposed CGAN REM model possesses an improved accuracy in predicting the spectrum activities within the television spectrum which is considered appropriate for DSS technology. This study also revealed that 41 channels within TV-band in Covenant University are totally unoccupied.https://ieeexplore.ieee.org/document/10813343/Conditional generative adversarial network (CGAN)dynamic spectrum sharing (DSS)radio environment map (REM)received signal strength (RSS)television white spaces (TVWS) |
| spellingShingle | Oluwatobi Emmanuel Dare Kennedy Okokpujie Emmanuel Adetiba Olabode Idowu-Bismark Abdultaofeek Abayomi Raymond Jules Kala Emmanuel Owolabi Udeme Christopher Ukpong Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment Mapping IEEE Access Conditional generative adversarial network (CGAN) dynamic spectrum sharing (DSS) radio environment map (REM) received signal strength (RSS) television white spaces (TVWS) |
| title | Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment Mapping |
| title_full | Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment Mapping |
| title_fullStr | Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment Mapping |
| title_full_unstemmed | Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment Mapping |
| title_short | Development of a Conditional Generative Adversarial Network Model for Television Spectrum Radio Environment Mapping |
| title_sort | development of a conditional generative adversarial network model for television spectrum radio environment mapping |
| topic | Conditional generative adversarial network (CGAN) dynamic spectrum sharing (DSS) radio environment map (REM) received signal strength (RSS) television white spaces (TVWS) |
| url | https://ieeexplore.ieee.org/document/10813343/ |
| work_keys_str_mv | AT oluwatobiemmanueldare developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping AT kennedyokokpujie developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping AT emmanueladetiba developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping AT olabodeidowubismark developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping AT abdultaofeekabayomi developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping AT raymondjuleskala developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping AT emmanuelowolabi developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping AT udemechristopherukpong developmentofaconditionalgenerativeadversarialnetworkmodelfortelevisionspectrumradioenvironmentmapping |